For code-elements, I've always been interested in the idea of
maintaining human-meaningful names by "dodging" name collisions
dynamically.  For the most part, prefixes combined with our desire for
slim images have carried us this far, but one should expect a
name-collision eventually.  We can do what Java did, "use huge
prefixes" but then there's still the potential for a collision.  When
that happens, why shouldn't a dynamic system dynamically adjust it?

This is just an idea for getting modules to co-exist in a Smalltalk
image, it is completely independent of module authentication..


On 5/11/07, Tony Garnock-Jones <tonyg at lshift.net> wrote:
> J J wrote:
> > Perhaps that is the way to go, we'll see, but one must keep in mind that
> > there is a slider bar: on one side is "user friendly-ness" and on the
> > other side is "security".  The more secure you are, the less user
> > friendly.  The more user friendly, the less secure.
>
> I'm not sure it's quite so binary.
>
> There's an idea called "Zooko's Triangle",
> http://en.wikipedia.org/wiki/Zooko's_triangle, that states that names
> can be any *two* of
>
>  - human-meaningful/friendly/memorable
>  - decentralised
>  - secure/unique
>
> but never all three at once.
>
> The situation we have within a Squeak image today is a *centralised*
> human-meaningful unique naming system: classes have unique names in a
> central, system-wide directory.
>
> The situation we have within the Squeak community at large is a little
> different: a decentralised, human-meaningful, but *non-unique* naming
> system, where anybody at all can build their own MyClass, completely
> different to everybody else's.
>
> Systems like Craig's naiad, if I understand correctly, are decentralised
> and unique, and sacrifice a certain degree of human-meaningfulness. This
> strikes me as the way to go, because tools can help bridge the gap to a
> system just as easy to use as the current centralised class dictionary.
>
> For an excellent introduction to the idea of "petnames",
> locally-meaningful names for securely-unique keys, please see
> http://www.skyhunter.com/marcs/petnames/IntroPetNames.html
>
> The "examples and near examples" section is particularly interesting -
> it's easy to imagine Squeak's tooling slightly refactored to provide
> outstanding support for a petname system for classes.
>
> More references:
>
> http://zooko.com/distnames.html
> http://shirky.com/writings/domain_names.html
>
> Regards,
>   Tony
> --
>  [][][] Tony Garnock-Jones     | Mob: +44 (0)7905 974 211
>    [][] LShift Ltd             | Tel: +44 (0)20 7729 7060
>  []  [] http://www.lshift.net/ | Email: tonyg at lshift.net
>
>